Factor VII assay performance: an analysis of the North American Specialized Coagulation Laboratory Association proficiency testing results

Authors


Correspondence:

Elizabeth M. Van Cott, Coagulation Laboratory, Massachusetts General Hospital GRJ235, 55 Fruit Street, Boston, MA 02114, USA.

Tel.: 617 726 9468;

Fax: 617 726 3256;

Summary

The performance of factor VII (FVII) assays currently used by clinical laboratories was examined in North American Specialized Coagulation Laboratory Association (NASCOLA) proficiency tests. Data from 12 surveys conducted between 2008 and 2010, involving 20 unique specimens plus four repeat-tested specimens, were analyzed. The number of laboratories per survey was 49–54 with a total of 1224 responses. Numerous reagent/instrument combinations were used. For FVII > 80 or <40 U/dL, 99.5% of results (859/863) were correctly classified by laboratories as normal/abnormal. Classification of specimens with 40–73 U/dL FVII was heterogeneous. Interlaboratory precision was better for normal specimens (coefficient of variation (CV) 10.7%) than for FVII<20 U/dL (CV 33.1%), with a mean CV of 17.2% per specimen. Intralaboratory precision for repeated specimens demonstrated no significant difference between the paired survey results (mean absolute difference 2.5–5.0 U/dL). For specimens with FVII >50 U/dL, among commonly used methods, one thromboplastin and one calibrator produced results 5–6 U/dL higher and another thromboplastin and calibrator produced results 5–6 U/dL lower than all other methods, and human thromboplastin differed from rabbit by +7.6 U/dL. Preliminary evidence suggests these differences could be due to the calibrator. For FVII <50 U/dL, differences among the commonly used reagents and calibrators were generally not significant.

Introduction

Factor VII is a serine protease that forms a complex with tissue factor to initiate the extrinsic pathway of coagulation by activating factor X [1]. Deficiencies of factor VII are associated with an increased risk of bleeding. Hereditary factor VII deficiencies are a rare cause of a bleeding disorder. Acquired decreases in factor VII are very common, caused most often by decreased hepatic synthesis, vitamin K deficiency, warfarin, or disseminated intravascular coagulation (DIC). A factor VIIa concentrate is commercially available, and in some countries, 4-factor prothrombin complex concentrates are available that include factor VII.

For patients with hereditary deficiencies, data from a review of the literature and database registries reported that the association between factor VII levels and clinical severity is not strong, but <10% activity was associated with major spontaneous bleeds [2-4]. The most common symptoms were prolonged bleeding provoked by injury and mucosal bleeding. Hematoma, hemarthrosis, and gastrointestinal bleeding also occurred, and a few severe cases experienced intracranial and umbilical cord bleeding [2]. Factor VII levels >25% were associated with asymptomatic individuals [5]. Factor VII <19% was associated with grade I bleeding (bleeding provoked by trauma or antiplatelet or anticoagulant therapy), <13% was associated with grade II bleeding (spontaneous minor bleeding), and <8% with grade III bleeding (spontaneous major bleeding) [5]. Surgical bleeding occurs in 15–24% of individuals [3, 6].

Databases have also attempted to address the question as to whether heterozygous deficient patients have a bleeding tendency, as well as homozygous individuals. When defining heterozygous deficiencies as those with ≥20 U/dL factor VII, a database reported that among 88 heterozygous deficient patients, 36% experienced hemorrhage (most commonly skin and mucus membrane bleeding), and most episodes were spontaneous (61%) [7].

For these reasons, the ability of laboratories to generate accurate determinations of factor VII levels is important in the care of patients with bleeding, for diagnostic and therapeutic purposes. In this study, we sought to assess the performance of factor VII assays among clinical laboratories in North America.

Methods

Proficiency surveys

Primary data were obtained from all 12 North American Specialized Coagulation Laboratory Association (NASCOLA) proficiency testing surveys conducted from 2008 to 2010. The NASCOLA proficiency testing program distributes lyophilized plasma samples to participating laboratories and is part of the External Quality Control of Diagnostic Assays and Tests (ECAT) Foundation proficiency testing program. Samples within each survey are from the same patients or control materials, and thus, all laboratories test identical specimens. Control materials were from Technoclone (Vienna, Austria) and Siemens (Munich, Germany). Laboratories do not know the expected results, and they process and test the proficiency test samples according to their clinical protocols used for patient specimens.

As shown in Table 1, among the 24 specimens used in this study, five specimens had <20 U/dL factor VII, five had 20–<40 U/dL factor VII, six had 40–<60 U/dL factor VII, one had 73 U/dL factor VII, and seven specimens had >80 U/dL factor VII. Four of the specimens were tested on two surveys: 2009–2 and 2009–4 (all-laboratory mean of approximately 16 U/dL), 2009–1 and 2009–3 (all-laboratory mean of approximately 27 U/dL), 2009–3 and 2010–3 (all-laboratory mean of approximately 48 U/dL), and 2008–4 and 2009–1 (all-laboratory mean of approximately 49 U/dL).

Table 1. Mean factor VII results and interlaboratory precision (CV) for the 24 specimens
SurveyMean FVII U/dL (range)CV (%)Description
  1. FVII, factor VII; CV, coefficient of variation.

  2. *, †, ‡, §Specimens were sent on two surveys.

2010–45.4 (2–19)53.0Patient with low FVII
2010–29.1 (5–16)27.8Sample with low FVII
2010–411.7 (3–23)34.3Commercial control
2009–415.3 (11–21)16.6Sample with low factors*
2009–216.4 (9–49)33.7Sample with low factors*
2008–321.5 (12–31)22.8Calibrant (INR approximately 3)
2009–327.0 (17–38)14.9Commercial control
2009–128.1 (18–50)16.8Commercial control
2010–332.0 (26–41)9.8Commercial control
2008–136.8 (25–46)14.3Calibrant (INR approximately 2)
2008–142.4 (33–51)9.7Borderline control
2010–347.4 (36–58)10.3Borderline control
2008–448.3 (32–67)15.9Abnormal control§
2009–349.2 (35–60)11.4Borderline control
2009–150.3 (27–69)18.3Abnormal control§
2008–259.2 (43–72)11.9Patient with factor X deficiency
2010–173.2 (53–94)16.7Patient with low factor X
2010–188.4 (62–103)9.6Normal control
2010–289.0 (64–132)12.8High factor II (160%)
2008–290.2 (67–108)9.6Patient with mildly high factor II
2008–491.7 (72–119)9.5Normal control
2009–499.2 (79–131)11.0High factor II
2008–3104.9 (77–132)9.4Patient with high factor II
2009–2110.3 (65–138)13.3Normal control

Data analysis and statistics

Coefficient of variation (CV) was calculated for each of the 24 samples to assess factor VII assay precision at various levels of factor VII. CV was also calculated for each thromboplastin reagent, deficient plasma, and calibrator that was used by more than 10 laboratories. As only one factor VII-deficient plasma was used by more than 10 laboratories on every survey, CV was also calculated for the next two most commonly used deficient plasmas.

The ability of laboratories to correctly categorize their results as normal or abnormal was assessed.

To analyze intralaboratory precision over time, a paired t-test was used for analysis of the four specimens that were tested on two separate proficiency surveys on two separate occasions 3, 6, or 12 months apart in laboratories that used the same methodology on both surveys.

Thromboplastin reagents, factor VII calibrators, and factor VII-deficient plasmas that were used by 10 or more laboratories were analyzed further. Unpaired t-test assuming unequal variances was used to assess for differences. P values <0.05 were considered statistically significant. A Bland–Altman plot was used to display observed differences.

Results

Between 2008 and 2010, 12 proficiency testing surveys were performed (four per year) with 20 unique specimens plus four of these 20 specimens were repeat-tested on a second survey, for a total of 24 specimens. The number of participant laboratories per proficiency survey ranged from 49 to 54 with a total of 1224 responses. All laboratories were in the United States or Canada. Among the 1224 responses, three outlier results were excluded from all analysis: 85 U/dL on survey 2010–4 (all-laboratory factor VII mean 11.5 U/dL), 88 U/dL on 2010–2 (all-laboratory factor VII mean 9.1 U/dL), and 72.2 U/dL on 2008–3 (all-laboratory factor VII mean 21.5 U/dL).

Table 1 shows the mean factor VII result and range for each specimen, as well as the interlaboratory precision.

Classification of results as normal vs. abnormal by laboratories

In addition to providing numerical results for factor VII, laboratories were asked to classify their results as normal, borderline normal, borderline abnormal, or abnormal, according to their own reference range (each laboratory determined their own criteria for classification). For the five specimens with very low factor VII (mean factor VII <20 U/dL), all 251 results were correctly classified by participants as abnormal or borderline abnormal. For the seven specimens with normal factor VII (mean factor VII >80 U/dL), 99.2% (351/354) of results were correctly classified as normal. Five specimens with a mean factor VII level of 20–40 U/dL were correctly classified as abnormal or borderline abnormal by all laboratories except 1 (99.6%, 257/258). Classification was more heterogeneous for the six specimens with borderline mean factor VII levels of 40–60 U/dL; the 308 responses were classified as normal by 21.8%, borderline normal or borderline abnormal by 14.9%, and abnormal by 63.3%.

As each laboratory establishes its own reference range, the definition of normal vs. abnormal was up to each individual laboratory. In some cases where the result was near the cutoff between normal and abnormal, it became evident that laboratory definitions of normal and abnormal differ. For example, in survey 2009–3, one specimen had a mean factor VII of 49.2 U/dL and four laboratories reported a factor VII of 50 U/dL; one of these laboratories classified the result as normal, two classified it as borderline normal, and one classified it as abnormal.

Interlaboratory precision (CV)

Table 2 shows that interlaboratory precision was better for specimens with >20 U/dL factor VII than for specimens with <20 U/dL factor VII. The mean CV% was 17.2% per specimen, ranging from 33.1% for very low (<20 U/dL) factor VII specimens to 10.7% for normal (>80 U/dL) factor VII specimens. CV for the most commonly used thromboplastin reagents, factor VII-deficient plasmas, and calibrators was similar to each other. For specimens >20 U/dL, the CV was lower for all reagents individually compared to the all-laboratory CV.

Table 2. Interlaboratory precision
Factor VII (U/dL)No. of specimensNo. of resultsCoefficient of variation (CV, %)
All methodsThromboplastinDeficient plasmaCalibrator
ABABCABC
  1. No., number.

  2. Thromboplastin A, Innovin.

  3. Thromboplastin B, Neoplastine CI Plus.

  4. Deficient plasma A, Siemens.

  5. Deficient plasma B, Stago.

  6. Deficient plasma C, Precision Biologic.

  7. Calibrator A, Siemens Standard Human Plasma.

  8. Calibrator B, Stago Unicalibrator.

  9. Calibrator C, Precision Biologic Normal Reference Plasma.

<20525133.127.236.327.842.531.628.236.433.8
20–<40525815.710.27.38.811.016.111.28.214.9
40–<60630812.98.07.68.09.912.49.38.210.8
60–<8015016.712.38.914.310.313.912.910.713.8
≥80735410.79.18.19.49.29.99.48.19.1
All24122117.213.013.713.016.716.513.814.216.1

Intralaboratory precision (CV) over time

Four specimens were tested on two separate surveys that were 3, 6, or 12 months apart. Results were compared from laboratories using the same method on both surveys. For the four specimens, there was no statistical difference between the means in laboratories using the same method on both surveys. For the individual laboratories, the mean difference for the for four paired results was −1.2, −1.1, −0.5, and +1.7 U/dL and the mean absolute difference for the four paired results for individual laboratories was +2.5, +3.0, +3.6, and +5.0 U/dL. Although the mean results for all laboratories overall were not significantly different, in a small number of individual laboratories their two results varied widely (up to 36 U/dL).

Reagent and calibrator differences

Numerous combinations of thromboplastin reagents, calibrators, factor VII-deficient plasmas, and equipment were used. For example, in 2008, 8–10 thromboplastin reagents, nine calibrators, nine factor VII-deficient plasmas, and 8–10 analyzers or analyzer families (a series of related analyzers from a manufacturer) were used, in varying combinations. In 2010, slightly fewer combinations were in use: seven thromboplastin reagents, seven calibrators, eight factor VII-deficient plasmas, and six analyzers or analyzer families.

Table 3 shows that the mean factor VII results were similar for the most commonly used thromboplastins, deficient plasmas, and calibrators when factor VII was low. More disparity appeared to be present for normal or borderline factor VII. Therefore, the more commonly used methods were analyzed further.

Table 3. Mean factor VII results
Factor VII (U/dL)No. of specimensNo. of resultsMean factor VII (U/dL)
All methodsThromboplastinDeficient plasmaCalibrator
ABABCABC
  1. No., number.

  2. Thromboplastin A, Innovin.

  3. Thromboplastin B, Neoplastine CI Plus.

  4. Deficient plasma A, Siemens.

  5. Deficient plasma B, Stago.

  6. Deficient plasma C, Precision Biologic.

  7. Calibrator A, Siemens Standard Human Plasma.

  8. Calibrator B, Stago Unicalibrator.

  9. Calibrator C, Precision Biologic Normal Reference Plasma.

<20525111.610.412.910.712.311.711.012.411.6
20–<40525829.129.330.429.931.228.130.030.327.9
40–<60630849.552.547.352.848.450.552.847.548.4
60–<8015073.283.060.182.759.774.183.460.573.1
≥80735496.2100.992.3101.292.498.5102.292.195.8
All24122152.054.350.354.650.652.755.050.251.3

Two thromboplastin reagents were used by 10 or more laboratories and could therefore be analyzed in greater detail: [A] Innovin (recombinant human, from Siemens Healthcare Diagnostics, Munich, Germany) was used by 20–28% of participating laboratories, and [B] Neoplastine CI Plus (rabbit brain, from Diagnostica Stago, Asnieres, France) was used by 29–35%.

For specimens with >50 U/dL factor VII, the mean difference between the two thromboplastins was 10.4 U/dL (thromboplastin A tended to produce higher levels with mean difference from the all-method mean +5.2 U/dL, and thromboplastin B tended to produce lower levels with mean difference from all-method mean −5.2 U/dL). Factor VII levels of 50–73 U/dL demonstrated the most difference: thromboplastin A + 6.5 U/dL above the all-laboratory mean and thromboplastin B −8.2 U/dL below the all-laboratory mean. This difference was statistically significant for all three surveys that had factor VII in this range (P < 0.02). At normal levels of factor VII (>80 U/dL), the difference was less pronounced (thromboplastin A mean difference from the all-laboratory mean +4.6 U/dL and thromboplastin B mean difference from all-laboratory mean −3.9 U/dL). For four of the seven proficiency surveys with factor VII >80 U/dL, the difference between thromboplastins A and B was statistically significant (P < 0.05). However, the difference from the mean (+4.6 and −3.9 U/dL respectively) is not usually clinically significant in this range of normal values, and the other three surveys showed no significant difference.

At low factor VII levels, the opposite pattern was observed, in that thromboplastin A tended to run slightly lower (−1.5 U/dL) and thromboplastin B slightly higher (+1.9 U/dL) than the all-laboratory mean in the six surveys with the lowest factor VII levels (<22 U/dL). For three of these six surveys, the difference between the two reagents was statistically significant (P < 0.04), but the slight difference is not usually expected to be clinically significant, and the other three surveys showed no significant difference.

Factor VII levels 22–50 U/dL had no predominant pattern emerge between the two reagents.

To determine whether the difference in reagents could be related to the calibrator, data were also analyzed by calibrator. Three calibrators were used by at least 10 laboratories and therefore, these could be analyzed further: [A] Siemens Standard Human Plasma (Siemens Healthcare Diagnostics), [B] Stago Unicalibrator (Diagnostica Stago), and [C] Precision Biologic Normal Reference Plasma (Precision Biologic, Dartmouth, Nova Scotia). The calibrators were used by a similar number of laboratories per survey: A 10–14 (20–27%), B 12–14 (25–28%), and C 12–15 (24–29%).

For calibrator A, 12 of the 24 samples showed no significant difference using calibrator A vs. all other calibrators, whereas the other 12 samples had significantly higher results using calibrator A vs. other calibrators (Figure 1a). Figure 1a shows that factor VII results from calibrator A tend to run higher than other calibrators for factor VII >50 U/dL. The mean difference in factor VII for calibrator A vs. all calibrators was +6.4 U/dL for samples with >50 U/dL factor VII and +0.7 U/dL for samples with <50 U/dL factor VII.

Figure 1.

Comparison of commonly used calibrators and thromboplastins. Bland–Altman plots showing the difference between the mean factor VII results obtained using (a) calibrator A, (b) calibrator B, or (c) calibrator C minus the mean results obtained using all calibrators (all-laboratory mean factor VII) and (d) the difference between rabbit and human thromboplastins against the all-laboratory mean factor VII. X indicates the results were significantly different. Circles indicate no significant difference.

For calibrator B, 12 of the 24 samples showed no significant difference using calibrator B vs. all other calibrators, whereas the other 12 samples showed either significantly higher or significantly lower results (Figure 1b). Figure 1b shows that calibrator B tended to run lower than other calibrators for factor VII >50 U/dL (mean difference −5.3 U/dL), but was similar to other calibrators for factor VII <50 U/dL (mean difference +0.7 U/dL).

For calibrator C, 22 of the 24 samples showed no significant difference using calibrator C vs. all other calibrators, and two samples were significantly lower using calibrator C vs. the other calibrators (Figure 1c). Thus, there is no apparent bias for calibrator C.

It is possible that the difference between the two most commonly used thromboplastin reagents might be at least partly due to the calibrator, because 70–80% of laboratories using thromboplastin A also use calibrator A. Thromboplastin A or calibrator A tended to run higher for factor VII >50 U/dL. None of the laboratories that use thromboplastin B used calibrator A, 71–86% used calibrator B, and thromboplastin B or calibrator B tended to give lower results than thromboplastin A for samples with >50 U/dL factor VII. As additional evidence that the reagent differences are due to the calibrator, one laboratory used thromboplastin A with calibrator B on four surveys with factor VII >50 U/dL, and the mean results for this laboratory on these four surveys (81.8 U/dL) matched the corresponding mean results for all thromboplastin B users (82.8 U/dL) and calibrator B users (82.5 U/dL) for these four surveys, and did not match the mean results for all thromboplastin A users (92.8 U/dL) nor calibrator A users (94.5 U/dL) for these four surveys.

The mean factor VII result using human thromboplastin was 52.9 U/dL and using rabbit thromboplastin was 50.5 U/dL. Figure 1d shows that rabbit thromboplastin results tended to run lower than human thromboplastin when factor VII was >50 and tended to run higher than human thromboplastin when factor VII was <50 U/dL. The mean difference in results of human vs. rabbit thromboplastin was +7.6 U/dL for specimens with factor VII >50 U/dL. For the reasons described above, this might be at least partly related to the choice of calibrator. Most laboratories (64.5%) using rabbit thromboplastin used calibrator B (vs. 0% used calibrator A), whereas 39.2% of laboratories using human thromboplastin used calibrator A (vs. 1.3% used calibrator B). As additional evidence that the difference may be due to the calibrator, one laboratory used human thromboplastin with calibrator B on four surveys with factor VII >50 U/dL, and the mean results for this laboratory on these surveys (81.8 U/dL) matched the corresponding mean results for all rabbit thromboplastin users (83.0 U/dL) and did not match the mean results for all human thromboplastin users (89.8 U/dL).

The factor VII-deficient plasmas most commonly used were [A] Siemens, [B] Stago, and [C] Precision Biologic. The deficient plasmas were used by 6–10 (12–20%) [A], 9–10 (17–20%) [B], and 16–21 (32–39%) [C] laboratories per survey. Only deficient plasma C was used by more than 10 laboratories on every survey and was analyzed further. Deficient plasma C tended to produce results near the all-laboratory mean (mean difference per specimen from the all-laboratory mean was +0.8 U/dL, and only three of 24 specimens were significantly different vs. all other calibrators). At factor VII levels >50 U/dL, deficient plasma A tended to produce higher results (mean difference +5.5 U/dL) and deficient plasma B tended to produce lower results (mean difference −4.9 U/dL). The differences among the deficient plasmas are similar to the results of the calibrators and thromboplastins most commonly used with the deficient plasmas (e.g., deficient plasma A is most commonly used with thromboplastin A and calibrator A), and thus, the differences may reflect the choice of calibrator and thromboplastin.

Discussion

This study found that interlaboratory precision is better for factor VII >20 U/dL (i.e., >20% factor VII) than for factor VII <20 U/dL and that most laboratories performed consistently over time. CV for the most commonly used reagents, deficient plasmas, and calibrators was similar to each other and to the all-laboratory CV. Almost all laboratories correctly classified their results as normal or abnormal when factor VII was >80 or <40 U/dL. Classification was more heterogeneous when factor VII was 40–73 U/dL.

The two most commonly used thromboplastin reagents showed some differences that are not likely to be clinically significant, except possibly in the range of 50–73 U/dL factor VII where the greatest difference between the two reagents was seen. In this range, thromboplastin A (Innovin, a human thromboplastin) was, on average, +6.5 U/dL above the all-laboratory mean and thromboplastin B (Neoplastine CI Plus, a rabbit thromboplastin) was −8.2 U/dL below the all-laboratory mean. We found evidence that differences between these two most commonly used thromboplastin reagents could be due to the calibrator rather than the reagent itself, although further study is needed for confirmation. This raises the suggestion that proficiency testing programs might need to grade participant results based on the calibrator or combination of calibrator and reagent, rather than the reagent alone. Currently, proficiency programs typically grade based on reagent.

Calibrator C had slightly higher CV than calibrators A and B (Table 2), but this could be because calibrator C tended to be used with a variety of thromboplastins and deficient plasmas, whereas calibrator A was used mainly with thromboplastin A and deficient plasma A, and calibrator B was used mainly with thromboplastin B and deficient plasma B.

In conclusion, factor VII assays generally performed well in terms of interlaboratory precision for normal factor VII levels and classification of patients as normal or abnormal for factor VII >80 or <40 U/dL. Not surprisingly, classification was more heterogeneous for mid-range factor VII levels 40–73 U/dL, a range that spans the typical lower limit of most laboratories' reference range. Areas where factor VII testing could improve include interlaboratory precision for factor VII <20 U/dL and more consistency among the different thromboplastins and calibrators for factor VII >50 U/dL.

Acknowledgements

EMVC and NDZ wrote the manuscript and were the primary investigators. MLK and PM contributed the data from ECAT. All authors reviewed and analyzed the data, reviewed the manuscript, and approved the final version.

Ancillary